Neuromorphic computers hold the potential to vastly improve the speed and efficiency of a wide range of computational kernels with their asynchronous, compute-memory co-located, spatially distributed, and scalable nature. However, performance models that are simple yet sufficiently expressive to predict runtime on actual neuromorphic hardware are lacking, posing a challenge for researchers and developers who strive to design fast algorithms and kernels. As breaking the memory bandwidth wall of conventional von-Neumann architectures is a primary neuromorphic advantage, modeling communication time is especially important. At the same time, modeling communication time is difficult, as complex congestion patterns arise in a heavily-loaded Network-on-Chip. In this work, we introduce the first max-affine lower-bound runtime model – a multi-dimensional roofline model – for Intel's Loihi 2 neuromorphic chip that quantitatively accounts for both compute and communication based on a suite of microbenchmarks. Despite being a lower-bound model, we observe a tight correspondence (Pearson correlation coefficient greater than or equal to 0.97) between our model's estimated runtime and the measured runtime on Loihi 2 for a neural network linear layer, i.e., matrix-vector multiplication, and for an example application, a Quadratic Unconstrained Binary Optimization solver. Furthermore, we derive analytical expressions for communication-bottlenecked runtime to study scalability of the linear layer, revealing an area-runtime tradeoff for different spatial workload configurations with linear to superliner runtime scaling in layer size with a variety of constant factors. Our max-affine runtime model helps empower the design of high-speed algorithms and kernels for Loihi 2.

Linear recurrent neural networks enable powerful long-range sequence modeling with constant memory usage and time-per-token during inference. These architectures hold promise for streaming applications at the edge, but deployment in resource-constrained environments requires hardware-aware optimizations to minimize latency and energy consumption. Unstructured sparsity offers a compelling solution-when accelerated by compatible hardware platforms. In this paper, we investigate the Pareto front of performance and efficiency across inference compute budgets. We find that highly sparse linear RNNs consistently achieve better efficiency-performance trade-offs than dense baselines, with 2x less compute and 36% less memory iso-accuracy. Our models achieve state-of-the-art results on a streaming audio denoising task. By quantizing our sparse models to fixed-point arithmetic and deploying them on the Intel Loihi 2 neuromorphic chip, we translate model compression into tangible gains of 42x lower latency and 149x lower energy consumption compared to a dense model on an edge GPU. Our findings showcase the transformative potential of unstructured sparsity, paving the way for highly efficient recurrent neural networks in real-world, resource-constrained environments. https://github.com/IntelLabs/SparseRNNs

The unsustainable rise in energy cost from increasingly capable deep learning systems spurs computer architecture innovation beyond conventional deep learning accelerators such as GPUs. However, a novel computer architecture presents a problem: much of deep learning research has been optimized for conventional computer architectures, and the extent to which modern deep learning models can unlock improved efficiency on a novel computer architecture is not well understood. In this work, we demonstrate for the first time that a State Space Model (SSM) can achieve substantial efficiency improvement when mapped to Loihi 2, a state-of-the-art neuromorphic research chip, versus a Jetson Orin Nano GPU (Jetson). Specifically, we benchmark our SSM on sMNIST, psMNIST, and sCIFAR online token-by-token inference and find approximately 1000x increased energy efficiency and 75x improved latency and throughput on Loihi 2 with a decrease in accuracy of less than one to three percentage points compared to the full precision implementation on Jetson. We comprehensively tailor our implementation to Loihi-specific features and constraints, such as the co-location of memory and compute as well as fixed precision arithmetic. Our results elucidate how SSMs meaningfully bridge conventional and neuromorphic hardware via their dual nature: SSMs can operate in an offline mode using convolution or scan, which is efficient on a GPU, or in an online mode as a recurrent network, which we show is efficient on Loihi 2. This work provides a foundation for performant sequence models on neuromorphic hardware, potentially unlocking substantial improvements in latency-sensitive or energy-limited online inference applications, such as speech enhancement or vision for robotic control.

Spiking Neural Networks are attracting increased attention as a more energy-efficient alternative to traditional Artificial Neural Networks for edge computing. Neuromorphic computing can significantly reduce energy requirements. Here, we present a complete pipeline: efficient event-based training of SNNs with synaptic delays on GPUs and deployment on Intel's Loihi 2 neuromorphic chip. We evaluate our approach on keyword recognition tasks using the Spiking Heidelberg Digits and Spiking Speech Commands datasets, demonstrating that our algorithm can enhance classification accuracy compared to architectures without delays. Our benchmarking indicates almost no accuracy loss between GPU and Loihi 2 implementations, while classification on Loihi 2 is up to 18x faster and uses 250x less energy than on an NVIDIA Jetson Orin Nano.

Neuromorphic computing shows promise for advancing computing efficiency and capabilities of AI applications using brain-inspired principles. However, the neuromorphic research field currently lacks standardized benchmarks, making it difficult to accurately measure technological advancements, compare performance with conventional methods, and identify promising future research directions. Prior neuromorphic computing benchmark efforts have not seen widespread adoption due to a lack of inclusive, actionable, and iterative benchmark design and guidelines. To address these shortcomings, we present NeuroBench: a benchmark framework for neuromorphic computing algorithms and systems. NeuroBench is a collaboratively-designed effort from an open community of researchers across industry and academia, aiming to provide a representative structure for standardizing the evaluation of neuromorphic approaches. The NeuroBench framework introduces a common set of tools and systematic methodology for inclusive benchmark measurement, delivering an objective reference framework for quantifying neuromorphic approaches in both hardware-independent (algorithm track) and hardware-dependent (system track) settings. In this article, we outline tasks and guidelines for benchmarks across multiple application domains, and present initial performance baselines across neuromorphic and conventional approaches for both benchmark tracks. NeuroBench is intended to continually expand its benchmarks and features to foster and track the progress made by the research community.

Loihi 2 is an asynchronous, brain-inspired research processor that generalizes several fundamental elements of neuromorphic architecture, such as stateful neuron models communicating with event-driven spikes, in order to address limitations of the first generation Loihi. Here we explore and characterize some of these generalizations, such as sigma-delta encapsulation, resonate-and-fire neurons, and integer-valued spikes, as applied to standard video, audio, and signal processing tasks. We find that these new neuromorphic approaches can provide orders of magnitude gains in combined efficiency and latency (energy-delay-product) for feed-forward and convolutional neural networks applied to video, audio denoising, and spectral transforms compared to state-of-the-art solutions.

The field of neuromorphic computing holds great promise in terms of advancing computing efficiency and capabilities by following brain-inspired principles. However, the rich diversity of techniques employed in neuromorphic research has resulted in a lack of clear standards for benchmarking, hindering effective evaluation of the advantages and strengths of neuromorphic methods compared to traditional deep-learning-based methods. This paper presents a collaborative effort, bringing together members from academia and the industry, to define benchmarks for neuromorphic computing: NeuroBench. The goals of NeuroBench are to be a collaborative, fair, and representative benchmark suite developed by the community, for the community. In this paper, we discuss the challenges associated with benchmarking neuromorphic solutions, and outline the key features of NeuroBench. We believe that NeuroBench will be a significant step towards defining standards that can unify the goals of neuromorphic computing and drive its technological progress. Please visit neurobench.ai for the latest updates on the benchmark tasks and metrics.

A critical enabler for progress in neuromorphic computing research is the ability to transparently evaluate different neuromorphic solutions on important tasks and to compare them to state-of-the-art conventional solutions. The Intel Neuromorphic Deep Noise Suppression Challenge (Intel N-DNS Challenge), inspired by the Microsoft DNS Challenge, tackles a ubiquitous and commercially relevant task: real-time audio denoising. Audio denoising is likely to reap the benefits of neuromorphic computing due to its low-bandwidth, temporal nature and its relevance for low-power devices. The Intel N-DNS Challenge consists of two tracks: a simulation-based algorithmic track to encourage algorithmic innovation, and a neuromorphic hardware (Loihi 2) track to rigorously evaluate solutions. For both tracks, we specify an evaluation methodology based on energy, latency, and resource consumption in addition to output audio quality. We make the Intel N-DNS Challenge dataset scripts and evaluation code freely accessible, encourage community participation with monetary prizes, and release a neuromorphic baseline solution which shows promising audio quality, high power efficiency, and low resource consumption when compared to Microsoft NsNet2 and a proprietary Intel denoising model used in production. We hope the Intel N-DNS Challenge will hasten innovation in neuromorphic algorithms research, especially in the area of training tools and methods for real-time signal processing. We expect the winners of the challenge will demonstrate that for problems like audio denoising, significant gains in power and resources can be realized on neuromorphic devices available today compared to conventional state-of-the-art solutions.

Neural circuits consist of many noisy, slow components, with individual neurons subject to ion channel noise, axonal propagation delays, and unreliable and slow synaptic transmission. This raises a fundamental question: how can reliable computation emerge from such unreliable components? A classic strategy is to simply average over a population of N weakly-coupled neurons to achieve errors that scale as 1/N. But more interestingly, recent work has introduced networks of leaky integrate-and-fire (LIF) neurons that achieve coding errors that scale superclassically as 1/N by combining the principles of predictive coding and fast and tight inhibitory-excitatory balance. However, spike transmission delays preclude such fast inhibition, and computational studies have observed that such delays can cause pathological synchronization that in turn destroys superclassical coding performance. Intriguingly, it has also been observed in simulations that noise can actually improve coding performance, and that there exists some optimal level of noise that minimizes coding error. However, we lack a quantitative theory that describes this fascinating interplay between delays, noise and neural coding performance in spiking networks. In this work, we elucidate the mechanisms underpinning this beneficial role of noise by deriving analytical expressions for coding error as a function of spike propagation delay and noise levels in predictive coding tight-balance networks of LIF neurons. Furthermore, we compute the minimal coding error and the associated optimal noise level, finding that they grow as power-laws with the delay. Our analysis reveals quantitatively how optimal levels of noise can rescue neural coding performance in spiking neural networks with delays by preventing the build up of pathological synchrony without overwhelming the overall spiking dynamics. This analysis can serve as a foundation for the further study of precise computation in the presence of noise and delays in efficient spiking neural circuits.

Analogue memory-based deep neural networks provide energy-efficiency and per-area throughput gains relative to state-of-the-art digital counterparts such as graphics processing units. Recent advances focus largely on hardware-aware algorithmic training and improvements to circuits, architectures, and memory devices. Optimal translation of software-trained weights into analogue hardware weights-given the plethora of complex memory non-idealities-represents an equally important task. We report a generalised computational framework that automates the crafting of complex weight programming strategies to minimise accuracy degradations during inference, particularly over time. The framework is agnostic to network structure and generalises well across recurrent, convolutional, and transformer neural networks. As a highly flexible numerical heuristic, the approach accommodates arbitrary device-level complexity, making it potentially relevant for a variety of analogue memories. By quantifying the limit of achievable inference accuracy, it also enables analogue memory-based deep neural network accelerators to reach their full inference potential.

Biological neural networks face a formidable task: performing reliable computations in the face of intrinsic stochasticity in individual neurons, imprecisely specified synaptic connectivity, and nonnegligible delays in synaptic transmission. A common approach to combatting such biological heterogeneity involves averaging over large redundant networks of $N$ neurons resulting in coding errors that decrease classically as $1/\sqrt{N}$. Recent work demonstrated a novel mechanism whereby recurrent spiking networks could efficiently encode dynamic stimuli, achieving a superclassical scaling in which coding errors decrease as $1/N$. This specific mechanism involved two key ideas: predictive coding, and a tight balance, or cancellation between strong feedforward inputs and strong recurrent feedback. However, the theoretical principles governing the efficacy of balanced predictive coding and its robustness to noise, synaptic weight heterogeneity and communication delays remain poorly understood. To discover such principles, we introduce an analytically tractable model of balanced predictive coding, in which the degree of balance and the degree of weight disorder can be dissociated unlike in previous balanced network models, and we develop a mean field theory of coding accuracy. Overall, our work provides and solves a general theoretical framework for dissecting the differential contributions neural noise, synaptic disorder, chaos, synaptic delays, and balance to the fidelity of predictive neural codes, reveals the fundamental role that balance plays in achieving superclassical scaling, and unifies previously disparate models in theoretical neuroscience.

Simultaneously modeling source code and natural language has many exciting applications in automated software development and understanding. Pursuant to achieving such technology, we introduce PYMT5, the PYTHON method text-to-text transfer transformer, which is trained to translate between all pairs of PYTHON method feature combinations: a single model that can both predict whole methods from natural language documentation strings (docstrings) and summarize code into docstrings of any common style. We present an analysis and modeling effort of a large-scale parallel corpus of 26 million PYTHON methods and 7.7 million method-docstring pairs, demonstrating that for docstring and method generation, PYMT5 outperforms similarlysized auto-regressive language models (GPT2) which were English pre-trained or randomly initialized. On the CODE-SEARCHNET test set, our best model predicts 92.1% syntactically correct method bodies, achieved a BLEU score of 8.59 for method generation and 16.3 for docstring generation (summarization), and achieved a ROUGE-L F-score of 24.8 for method generation and 36.7 for docstring generation.